Supervised Learning

Supervised learning is the type of machine learning you use when you have labelled data: examples where you already know the correct answer. The model learns the relationship between inputs and labels from those examples, then predicts labels for new data. MLforSEO frames the choice simply: if you know the right output for at least some of your data, you can use supervised learning.

How it works

• Collect labelled examples, such as queries tagged with their search intent or reviews tagged as positive or negative.
• Split the data into a training set and a test set the model never sees during training.
• Train the model to map inputs to labels.
• Evaluate on the test set with metrics such as accuracy, precision, recall or F1.
• Predict labels for the rest of your data.

Supervised tasks come in two kinds. Classification predicts a category: binary (spam or not), multi-class (one of several intents) or multi-label (several tags at once). Regression predicts a number, such as next month’s traffic. Named entity recognition and sentiment analysis are also supervised tasks: the APIs you call were trained on large sets of labelled text.

How it’s used in SEO and marketing

• Search intent classification of keyword lists.
• Content and page-type classification across a site.
• Sentiment and entity analysis through pre-trained APIs.
• Spam, toxicity and moderation checks for user-generated content.
• Forecasting traffic or conversions with regression.
• Lead scoring and churn prediction from CRM data.

Example

A team labels 500 queries by intent by hand, trains a logistic regression classifier on them, and checks it on 100 held-out queries. If accuracy is acceptable, the model labels the remaining 50,000 queries in minutes. Where the model is unsure, those queries go back for human labelling and retraining, an approach known as active learning.

Limitations and things to watch

• Labels are expensive and slow to create, and inconsistent labels produce an inconsistent model.
• The model only knows the categories you trained it on and can drift as language and SERPs change.
• High accuracy on unbalanced data can mislead; check precision and recall per class.
• Pre-trained APIs save labelling work but use the provider’s categories, which may not match yours.
• When you have no labels, start with unsupervised learning or zero-shot models.

Learn it on MLforSEO

• The Characteristics of a Machine Learning Problem: A Task, Data and Solution Framework for SEOs
• What Machine Learning Classification Is, and Where It Fits in SEO
• Search-Intent Classification with Machine Learning: The Approaches Explained
• Complete Guide Understanding ML Classification for Marketers: Marketing Use Cases & Model Choices
• The Introduction to Machine Learning for SEO course